Reported `in_use_connections` grows beyond `pool_size` bounds

[pcorey]

I'm using finch 0.17.0 working with HTTP1 connections and a pool configuration like this:

pools: %{
  default: [
    size: 50,
    count: 1,
    pool_max_idle_time: :timer.seconds(120),
    conn_max_idle_time: :timer.seconds(120),
    start_pool_metrics?: true
  ]
}

Finch.get_pool_status is reporting an in_use_connections count that increases over time, eventually surpassing my pool size, which seems impossible.

I'm guessing it's because I'm configuring a conn_max_idle_time, and any connections that exceed that limit when being checked out are closed, but that closure isn't being recorded in the pool metrics.

[sneako]

Ah thank you! I see you have implemented your suggestion in your fork. Can you confirm that fixed this for you?

[pcorey]

@sneako I'm waiting on a deploy window to fully test my fix. I should know by the end of the week. If it looks good, I'll submit a PR!

[pcorey]

@sneako After watching my fix in prod for a while, I'm noticing that in_use_connections is still increasing above the max pool size, and available_connections is falling well below zero. So, I think I'm missing other cases. I'll put some more attention towards this, but do you have any ideas of what might be missing?

[sneako]

@sneako After watching my fix in prod for a while, I'm noticing that in_use_connections is still increasing above the max pool size, and available_connections is falling well below zero. So, I think I'm missing other cases. I'll put some more attention towards this, but do you have any ideas of what might be missing?

I don't have any ideas off the top of my head but I can try to take a look when I have some extra time.

Does it seem like there was any improvement at all with this change?

[pcorey]

@sneako Turns out I was just misreading the metrics. It looks like the available_connections has stayed stable around zero, and available_connections has stayed stable around my pool size. I think we're good! Submitting a PR shortly.